Stress-Driven Two-Phase Mixture Models

• Stress-Driven Two-Phase Mixture Models are continuum formulations where stress directly drives the interaction and evolution of phases, replacing strain-driven closures.
• They integrate local and nonlocal responses through convolution kernels and differential equations to ensure well-posed behavior in complex systems.
• Applications span nanotechnology, geodynamics, granular flows, and debris transport, offering robust frameworks for modeling stress-induced phase transitions.

Stress-driven two-phase mixture models are continuum formulations in which the interaction of two phases, two constitutive branches, or two kinematic regimes is governed directly by stress, stress resultants, collision pressure, capillarity stress, or internal stress variables rather than by a purely strain-driven closure. In the structural-mechanics literature, the phrase denotes a convex local/nonlocal constitutive mixture in which nonlocal deformation measures are generated by convolution of local source fields with an attenuation kernel (Vaccaro et al., 2021). In multiphase flow, suspension, granular, and yield-stress settings, it denotes models in which momentum balance, phase separation, jamming, or phase transition is driven by internal stress tensors, capillarity, effective stress, or collision pressure (Cheng et al., 29 Sep 2025, Ahnert et al., 2017). Across these uses, the common feature is that stress is not merely an output of the model; it is a primary state variable or driving field for constitutive mixing, interfacial transmission, or phase evolution.

1. Conceptual scope and defining structures

The designation “two-phase mixture” is not unique across the literature. In stress-driven integral elasticity for beams, the two phases are a local phase and a nonlocal phase combined through a convex mixture parameter. In granular solids, the two phases are a marginally stable isostatic phase and an over-connected elastic phase. In yield-stress fluids and suspensions, the phases are solid-like and fluid-like constituents with distinct mass, momentum, and entropy balances. In diffuse-interface models, the phase distribution is represented by an order parameter such as $\varphi\in[-1,1]$, while stress terms couple that field back into momentum balance (Vaccaro et al., 2021, Blumenfeld, 2023, Peshkov et al., 2013).

The same diversity appears in the meaning of “stress-driven.” In the beam models of Barretta and coauthors, the nonlocal constitutive response is driven by statically admissible stress resultants, which avoids the paradoxes associated with strain-driven Eringen-type reductions on finite domains (Barretta et al., 2020). In viscoelastoplastic two-phase flows, the momentum equation is driven by a viscous contribution, an internal stress contribution $\eta(\varphi)S$, and capillarity, while the internal stress itself has its own evolution law with Zaremba–Jaumann rate and plastic subdifferential (Cheng et al., 29 Sep 2025). In granular composites, stress propagates through a hyperbolic isostatic phase and an elliptic elastic phase, so the two-phase character is encoded at the level of the PDE class itself (Blumenfeld, 2023).

A plausible implication is that the term identifies a family of models rather than a single canonical formalism. What unifies that family is the elevation of stress from a constitutive consequence to a constitutive driver.

2. Structural mechanics: local/nonlocal stress-driven mixtures

The most explicit constitutive use of the term appears in stress-driven two-phase integral elasticity for Timoshenko beams. For a planar curved beam with centerline $x(s)$, arc-length parameter $s\in[0,L]$, tangent $t(s)=x'(s)$, normal $n(s)=Rt(s)$, and curvature $c(s)$ defined by the Frenet formulas, the Timoshenko strain measures are

$$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$

Equilibrium is written in coordinate-free form as

$F'(s)+p(s)=0,\qquad M'(s)+T(s)+m(s)=0,$

or, after projection on the Frenet frame,

$$N'(s)-c(s)T(s)=-p(s)\cdot t(s),\qquad T'(s)+c(s)N(s)=-p(s)\cdot n(s),\qquad M'(s)+T(s)+m(s)=0.$$

The local curved-beam constitutive laws couple axial force and bending moment through curvature-dependent coefficients $\eta(\varphi)S$0 and $\eta(\varphi)S$1: $\eta(\varphi)S$2 These reduce to classical Timoshenko relations in the straight-beam limit $\eta(\varphi)S$3 (Vaccaro et al., 2021).

The stress-driven two-phase constitutive law then takes the form

$\eta(\varphi)S$4

where $\eta(\varphi)S$5 is the nonlocal deformation vector, $\eta(\varphi)S$6 is the local source vector, and the kernel is the bi-exponential

$\eta(\varphi)S$7

Here $\eta(\varphi)S$8 weights the local versus nonlocal phase, and $\eta(\varphi)S$9 is the internal length. For this kernel, the integral model is equivalent to the differential problem

$x(s)$0

with constitutive boundary conditions

$x(s)$1

This equivalence is central: the model is purely local for $x(s)$2, purely nonlocal for $x(s)$3, and remains well-posed because the source fields are computed from equilibrated resultants (Vaccaro et al., 2021).

The straight-beam precursor adopts the same stress-driven local/nonlocal mixture for Timoshenko nano-beams, with curvature and shear strain expressed as stress-driven convolutions of $x(s)$4 and $x(s)$5, respectively (Barretta et al., 2020). That formulation likewise leads to Helmholtz-type differential equations and constitutive boundary conditions for bending and shear. The literature emphasizes the contrast with strain-driven Eringen-like models, whose constitutive boundary conditions may be incompatible with finite-domain mechanical boundary data (Barretta et al., 2020).

In the curved-beam setting, the parametric effects are explicit. For a SiC nanobeam cantilever under a point force $x(s)$6, the tip transverse displacement $x(s)$7 in units of $x(s)$8 decreases from $x(s)$9 at $s\in[0,L]$0, $s\in[0,L]$1 to $s\in[0,L]$2 at $s\in[0,L]$3, $s\in[0,L]$4, while increasing to $s\in[0,L]$5 at $s\in[0,L]$6, $s\in[0,L]$7 (Vaccaro et al., 2021). The reported trend is therefore stiffening with increasing $s\in[0,L]$8 and softening with increasing $s\in[0,L]$9.

3. Diffuse-interface, two-pressure, and filtered flow formulations

In two-phase flow, stress-driven formulations typically place stress directly inside the momentum equation and often promote additional stress variables to dynamical unknowns. A representative example is the viscoelastoplastic Cahn–Hilliard model for two-phase geodynamic flow, in which incompressibility is enforced by $t(s)=x'(s)$0 and the momentum balance is

$t(s)=x'(s)$1

with

$t(s)=x'(s)$2

The phase field satisfies

$t(s)=x'(s)$3

and the internal stress obeys

$t(s)=x'(s)$4

The model admits Leray–Hopf-type weak solutions for $t(s)=x'(s)$5 and dissipative solutions in the vanishing stress-diffusion limit $t(s)=x'(s)$6 (Cheng et al., 29 Sep 2025).

A different but related stress-driven closure appears in quasi-incompressible diffuse-interface Navier–Stokes–Cahn–Hilliard theory for fluids with different densities. There the mixture velocity is nonsolenoidal in interfacial regions,

$t(s)=x'(s)$7

and momentum is written as

$t(s)=x'(s)$8

Here the capillary or Korteweg contribution is represented in the stress-driven form $t(s)=x'(s)$9, and the model is paired with a linear energy-stable time discretization that preserves mass and satisfies discrete energy dissipation independent of the time-step size, provided that the mixture density remains positive (Roudbari et al., 2016).

The nonequilibrium two-pressure SHTC model extends this logic by allowing $n(s)=Rt(s)$0, introducing an interfacial microinertia variable $n(s)=Rt(s)$1 and an interface gradient variable $n(s)=Rt(s)$2, and choosing the energy

$n(s)=Rt(s)$3

The resulting capillary stress is

$n(s)=Rt(s)$4

and the pressure imbalance enters through

$n(s)=Rt(s)$5

Formal asymptotics recover a diffuse-interface Young–Laplace-type relation in the fast-relaxation limit (Peshkov et al., 2023).

A further constitutive refinement appears in filtered one-fluid formulations of incompressible two-phase flow. Recognizing that fixed-grid solvers imply spatial filtering, the viscous stress tensor can no longer be represented by a single isotropic viscosity in interface-straddling control volumes. The rational anisotropic law introduces two distinct coefficients,

$n(s)=Rt(s)$6

and builds the filtered stress from the filtered strain-rate tensor and interface normal $n(s)=Rt(s)$7 (Magnaudet et al., 11 Apr 2025). In a three-dimensional benchmark of viscous buoyancy-driven exchange flow in a closed vertical pipe, the anisotropic model is reported as the only one capable of predicting correctly the evolution of the fronts of the ascending and descending fingers at a reasonable computational cost (Magnaudet et al., 11 Apr 2025).

4. Granular, suspension, sediment, and debris-flow models

In granular solids, Blumenfeld formulates stress transmission itself as a two-phase composite problem. The marginally stable phase obeys the isostatic closure

$n(s)=Rt(s)$8

which yields hyperbolic stress equations and characteristic stress chains, whereas the over-connected phase obeys linear elasticity,

$n(s)=Rt(s)$9

which is elliptic (Blumenfeld, 2023). The interface is not handled by standard displacement continuity alone; instead, the stress delivered by the isostatic side prescribes a deformation on the elastic side through

$c(s)$0

For finite regions, the stability parameter

$c(s)$1

classifies the region as fluid when $c(s)$2, marginally stable when $c(s)$3, and solid when $c(s)$4 (Blumenfeld, 2023).

Concentrated suspension theory uses collision pressure as the stress variable that controls yielding and jamming. In the yielded solid phase,

$c(s)$5

while in the jammed phase

$c(s)$6

In plane Poiseuille flow this produces a finite-width jammed core, and a matched-asymptotic drift-flux reduction captures the appearance and evolution of the jammed region (Ahnert et al., 2017).

The fluid-sediment model of Dunatunga and Kamrin organizes two-phase coupling around effective stress, dilatancy, and Darcy-like drag. The mixture stress is

$c(s)$7

with barotropic pore pressure

$c(s)$8

drag

$c(s)$9

and dilatancy law

$$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$0

The model is designed to span dilute suspensions, dense submerged granular flows, pure fluid, and dry grains (Baumgarten et al., 2018).

Debris-flow phase separation is formulated even more explicitly as a stress-driven flux. Relative slip is driven by Coulomb friction, pressure-gradient, buoyancy, topographic, and shear-rate forces, and the resulting solid and fluid separation fluxes are

$$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$1

with analogous cross-slope terms. In a one-dimensional reduction, the solid fraction obeys an advection–anti-diffusion equation with positive anti-diffusion coefficient $$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$2, so that $$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$3 diffuses “up-hill” against its gradient (Pudasaini et al., 2016). Simulations reported for a $$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$4 slope show a solid-rich frontal surge head, lateral levees, and a fluid-rich tail (Pudasaini et al., 2016).

5. Thermodynamic, variational, and Hamiltonian structure

A notable feature of stress-driven two-phase mixture models is that many are built to satisfy mechanics and thermodynamics simultaneously. In the Euler-structured model for yield-stress fluids, the two-phase continuum is described by local conservation laws of Godunov type, with stress-induced solid–fluid phase transition entering through dissipative source terms in the mass fraction and strain-dissipation equations (Peshkov et al., 2013). The Lagrangian formulation possesses full Godunov structure, and the Eulerian formulation retains energy conservation and thermodynamic consistency, even though gauge constraints complicate full symmetrization (Peshkov et al., 2013).

The energetic variational approach to diffuse-interface multiphase materials derives the reversible stress directly from the mixing energy through the principle of virtual work. For the energy density $$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$5, the elastic stress is

$$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$6

which, for the specific three-phase model, becomes

$$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$7

The total energy obeys

$$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$8

(Brannick et al., 2014).

The viscoelastoplastic Cahn–Hilliard model uses a total free energy combining kinetic, elastic, and phase-field terms and derives a relative energy inequality for dissipative solutions (Cheng et al., 29 Sep 2025). The quasi-incompressible NS–CH model derives its constitutive choices through mixture theory and the Coleman–Noll procedure, then proves continuous and discrete energy dissipation (Roudbari et al., 2016). The SHTC two-pressure model supplements this by giving the reversible subsystem both variational and Hamiltonian structures, with entropy production

$$\varepsilon(s)=u'(s)\cdot t(s),\qquad \gamma(s)=u'(s)\cdot n(s)-\varphi(s),\qquad \chi(s)=\varphi'(s).$$9

(Peshkov et al., 2023).

This suggests that thermodynamic compatibility is not incidental in this literature. It is often the mechanism by which stress-driven couplings are made mathematically admissible.

6. Interfaces, boundary conditions, well-posedness, and limitations

Stress-driven models frequently differ from more classical mixture theories in the role played by nonstandard boundary and interface conditions. In stress-driven beam elasticity, the constitutive boundary conditions are neither optional nor cosmetic; they are required by the equivalence between the integral and differential forms and complement the static and kinematic boundary conditions to secure well-posedness (Vaccaro et al., 2021). The straight-beam version makes the same point for bending and shear fields in Timoshenko nano-beams (Barretta et al., 2020).

In granular two-phase composites, the difficulty is not a higher-order constitutive boundary condition but a unique transmission constraint between phases obeying different PDE classes. The isostatic side carries stress without a strain-based constitutive law, so transmission into the elastic side requires prescribed interfacial deformation chosen to match the stress delivered by the isostatic phase (Blumenfeld, 2023). In filtered one-fluid flow, the interface induces constitutive anisotropy through the filtered normal $F'(s)+p(s)=0,\qquad M'(s)+T(s)+m(s)=0,$0, and the failure to distinguish normal and shear stress transmission leads to systematic errors in arithmetic or harmonic isotropic blending laws (Magnaudet et al., 11 Apr 2025).

Well-posedness remains model-dependent. The stress-driven structural mixtures are explicitly presented as a remedy to the ill-posedness of strain-driven Eringen-type finite-domain problems (Vaccaro et al., 2021). The viscoelastoplastic two-phase flow has global weak solutions only in the stress-diffusion regularized case and requires dissipative solutions for the non-regularized limit (Cheng et al., 29 Sep 2025). The nonequilibrium two-pressure SHTC model retains energy conservation and thermodynamic compatibility, but the paper reports that global symmetric hyperbolicity is not guaranteed for common equations of state because the internal energy part may be non-convex (Peshkov et al., 2023). The debris-flow formulation is depth-averaged and parameterizes vertical shear rather than resolving it, which limits the detailed vertical structure it can represent (Pudasaini et al., 2016).

A common misconception is that “stress-driven two-phase mixture model” denotes a single standard constitutive form. The literature instead uses the phrase for several mathematically distinct constructions unified only by the constitutive primacy of stress.

7. Applications, significance, and open directions

The application range is unusually broad. In nanotechnology, stress-driven two-phase integral elasticity is proposed for the design and optimization of curved micro- and nano-beam sensors and actuators, with closed-form or computationally light solutions suitable for parametric studies (Vaccaro et al., 2021). In geodynamics, viscoelastoplastic Cahn–Hilliard models target lithospheric and mantle materials on geological time scales (Cheng et al., 29 Sep 2025). In granular matter, the two-phase composite picture is intended to extend stress theory beyond marginal stability and the yield threshold (Blumenfeld, 2023). In multiphase CFD, the anisotropic filtered viscous stress law is benchmarked against a three-dimensional exchange-flow experiment with large viscosity contrast (Magnaudet et al., 11 Apr 2025). In debris-flow hazard assessment, explicitly evolving phase composition alters impact-pressure estimates and explains solid-rich frontal surges and levee formation (Pudasaini et al., 2016).

The major open directions are equally diverse. In viscoelastoplastic flow, uniqueness and higher regularity remain delicate because of three-dimensional Navier–Stokes-like structure and non-smooth plasticity (Cheng et al., 29 Sep 2025). In granular composites, predicting how the fabric tensor $F'(s)+p(s)=0,\qquad M'(s)+T(s)+m(s)=0,$1 reorganizes under changing boundary stress remains open (Blumenfeld, 2023). In nonequilibrium two-pressure flow, extensions to viscosity, heat conduction, and phase change remain natural next steps within the SHTC framework (Peshkov et al., 2023). In debris and sediment transport, calibration of drag, dilatancy, and separation parameters remains central for quantitative prediction (Pudasaini et al., 2016, Baumgarten et al., 2018).

Across these domains, stress-driven two-phase mixture models serve less as a single theory than as a structural principle: two-phase behavior is represented by constitutive branches, stress carriers, or phase fractions whose evolution is determined by stress itself. That principle has produced well-posed nonlocal beam theories, stress-evolving diffuse-interface flows, mixed hyperbolic–elliptic granular composites, collision-pressure-induced jamming models, and phase-separating debris-flow descriptions, each with its own variables and PDE structure but all organized around the same constitutive inversion of cause and effect.